Hi,
\nI am pretty new to kaggle. I submitted the model and everything ran fine. But I don't see any score for the submission. Any help. Am I doing it the right way.
I made my notebook public: https://www.kaggle.com/code/dhinkris/baselinesubmission/notebook
\nThank you.
", "messages": [ { "id": 2087584, "postDate": "2023-01-05T17:30:46.570Z", "content": "Hi,
\nI am pretty new to kaggle. I submitted the model and everything ran fine. But I don't see any score for the submission. Any help. Am I doing it the right way.
I made my notebook public: https://www.kaggle.com/code/dhinkris/baselinesubmission/notebook
\nThank you.
", "rawMarkdown": "Hi,\nI am pretty new to kaggle. I submitted the model and everything ran fine. But I don't see any score for the submission. Any help. Am I doing it the right way.\n\nI made my notebook public: https://www.kaggle.com/code/dhinkris/baselinesubmission/notebook\n\nThank you.", "votes": 3 }, { "id": 2092907, "postDate": "2023-01-09T17:27:58.327Z", "content": "Version 4 of the below notebook is a successful submission.
\nFew things to note while submitting is,
\ni. make sure you are reading the dcm images directly from the /test-image/ folder.
\nii. run inference in batches or one by one. This avoids OOM error.
\niii. store any temporary results in /kaggle/tmp/ and make sure to create before accessing.
\niv. make sure to run all training preprocessing steps from dicom read to prediction.
\niv. There maybe more than one image per patient for L or R. at the end it should one prediction per patient per laterality. Repetition can be avoided by dataframe.groupby('prediction_id').max().reset_index()
\nv. As mentioned in the Evaluation section, \" Submissions are evaluated using the probabilistic F1 score (pF1). This extension of the traditional F score accepts probabilities instead of binary classifications. You can find a Python implementation below.
Thank you. Good luck.
", "rawMarkdown": "Version 4 of the below notebook is a successful submission.\n\nFew things to note while submitting is, \ni. make sure you are reading the dcm images directly from the */test-image/* folder.\nii. run inference in batches or one by one. This avoids OOM error.\niii. store any temporary results in /kaggle/tmp/As mentioned in the Evaluation section, \" Submissions are evaluated using the probabilistic F1 score (pF1). This extension of the traditional F score accepts probabilities instead of binary classifications. You can find a Python implementation here: https://www.kaggle.com/code/sohier/probabilistic-f-score
", "rawMarkdown": "As mentioned in the Evaluation section, \" Submissions are evaluated using the probabilistic F1 score (pF1). This extension of the traditional F score accepts probabilities instead of binary classifications. You can find a Python implementation here: https://www.kaggle.com/code/sohier/probabilistic-f-score" }, { "id": 2090444, "postDate": "2023-01-07T10:41:58.793Z", "content": "How did you submit ? Did you use the \"Submit to competition\" menu on the right panel ?
", "rawMarkdown": "How did you submit ? Did you use the \"Submit to competition\" menu on the right panel ?", "replies": [ { "id": 2091815, "postDate": "2023-01-08T19:07:36.557Z", "content": "Yes, I used that. As @ellagale mentioned, I was using the png images from the test folder which doesn't have access to all the test images.
\nhere is the updated notebook:
\nhttps://www.kaggle.com/code/dhinkris/optimized-submission
Version 4 of the below notebook is a successful submission.
\nFew things to note while submitting is,
\ni. make sure you are reading the dcm images directly from the /test-image/ folder.
\nii. run inference in batches or one by one. This avoids OOM error.
\niii. store any temporary results in /kaggle/tmp/ and make sure to create before accessing.
\niv. make sure to run all training preprocessing steps from dicom read to prediction.
\niv. There maybe more than one image per patient for L or R. at the end it should one prediction per patient per laterality. Repetition can be avoided by dataframe.groupby('prediction_id').max().reset_index()
\nv. As mentioned in the Evaluation section, \" Submissions are evaluated using the probabilistic F1 score (pF1). This extension of the traditional F score accepts probabilities instead of binary classifications. You can find a Python implementation below.
Thank you. Good luck.
", "votes": 0, "replies": [] }, { "id": 2091817, "author_name": "dhinesh", "author_url": "", "post_date": "2023-01-08T19:09:58.353000", "content": "As mentioned in the Evaluation section, \" Submissions are evaluated using the probabilistic F1 score (pF1). This extension of the traditional F score accepts probabilities instead of binary classifications. You can find a Python implementation here: https://www.kaggle.com/code/sohier/probabilistic-f-score
", "votes": 0, "replies": [] }, { "id": 2090444, "author_name": "gguillard", "author_url": "", "post_date": "2023-01-07T10:41:58.793000", "content": "How did you submit ? Did you use the \"Submit to competition\" menu on the right panel ?
", "votes": 0, "replies": [ { "id": 2091815, "author_name": "dhinesh", "author_url": "", "post_date": "2023-01-08T19:07:36.557000", "content": "Yes, I used that. As @ellagale mentioned, I was using the png images from the test folder which doesn't have access to all the test images.
\nhere is the updated notebook:
\nhttps://www.kaggle.com/code/dhinkris/optimized-submission